Audio and video transcription

Choosing the right kind of transcript

Transcripts are not all the same, and the right one depends on what happens next. Clean verbatim removes filler words, false starts and repetition, so the text reads like edited speech. It suits interviews being turned into articles, meeting records and podcast show notes. Full verbatim keeps every "um", pause and interruption, which matters for academic research, legal review and user testing, where how something was said is part of the data.

Format matters too. A researcher wants timestamps at every speaker change to find quotes in the audio. A video editor wants SRT or VTT subtitle files with lines short enough to read on screen. A journalist wants plain text they can paste. Settling this before we start saves a second pass later.

Where accuracy is won or lost

General speech is the easy part. Errors cluster around names, brands, technical terms and acronyms, which are exactly the words that matter most in a business recording. A transcript that spells your product name three ways, or turns a drug name into a common word, has to be checked line by line before anyone can use it.

So we build a glossary before typing anything: speaker names, company and product names, industry terms and acronyms, from whatever notes or slides you can share. Each transcript then gets a second pass against the audio. Where a word truly cannot be made out, we mark it with a timestamp. A marked gap you can check in ten seconds is better than a confident guess you never notice.

• Record each speaker on their own microphone where possible
• Avoid rooms with hard walls and background music
• Send speaker names and a term list with the files